Re: [AD] file slices

[Jon Rafkind <workmin@xxxxxxxxxx>]

I'm guessing that you wanted the 'f' prefix to make people think about stdio operations but I think its not needed.

slice_close, slice_open, al_open_slice, etc

looks better to me.

On 07/29/2011 03:14 PM, Matthew Leverton wrote:

On Tue, Jul 19, 2011 at 7:09 PM, Peter Wang <novalazy@xxxxxxxxxx> wrote:

That still sounds complicated.  I prefer your previous suggestion
to remove any buffering from file slices.  Then file slices are
analogous to sub-bitmaps, and memfiles are analogous to memory bitmaps.

Attached is an implementation of unbuffered file slices. To run the
included test, you need to make test.tga that contains texture.tga
twice:

cat texture.tga texture.tga > test.tga

When using slices both can be loaded successfully, but without them
the second try fails. (I think the TGA loader doesn't seek to the end
of file when done.)

It would introduce one function:

ALLEGRO_FILE *al_fopen_slice(ALLEGRO_FILE *fp, size_t initial_size,
const char *mode);

It opens the slice at the current location of the file with an initial
size as specified by the user. Modes can be any combination of
[r]eadable, [w]ritable, and [e]xpandable.

If the file is not expandable, then writes and reads will be limited
to that initial size.

If the file is expandable, then the size will increase as data is read
or written. In such case, the initial_size would affect what SEEK_END
initially does.

There is no "until EOF" size, as the programmer can just do that
himself by passing al_fsize(fp) - al_ftell(fp) as initial size.

When closing the file, the parent is always moved to the end of the
slice's current size.

The following limitations exist:

* the underlying file must be random access (if reading)
* the underlying file must be open in regular write mode (if writing)
as the slice assumes that writes happen in place at the current
position.
* the parent cannot be used while the slice is open
* only one slice can be open per parent at a time, but slicing a slice
should theoretically work.

And of course there are some other edge cases that are left undefined,
like specifying a size that is larger than the parent, etc. But the
programmer should be well aware of what he is doing, so I don't think
that matters.

--
Matthew Leverton

------------------------------------------------------------------------------
Got Input?   Slashdot Needs You.
Take our quick survey online.  Come on, we don't ask for help often.
Plus, you'll get a chance to win $100 to spend on ThinkGeek.
http://p.sf.net/sfu/slashdot-survey